I have set up a RamDisk (2GB) and NCrunch filled it with builds. It then complains that there is not enough space on the disk. (I also sent a bug report about it).
Can you detect that you are running on a low capacity - high speed disk (<20GB, xMB/s) and assume it is a RamDrive. Then detect that there would be not enough space to run a build (given that you did at least one before). And clean up enough old builds to make sure you have enough space for the current run. This would allow NCrunch to run forever using the RamDrive, at the speed of light...
This is a good suggestion. Restricting the engine from building new workspaces when it needs them will reduce efficiency, but right now there isn't really any decent handling of situations where the drive runs out of space. The workspacing code was introduced before the discovery of using RAMdrives for workspaces, so it may need a little work here :) I've put it on the list of things to address in a future build.
Something you can try to reduce the workspace capacity needs is to reduce your MaxNumberOfProcessingThreads, as having a high number in this setting will result in the engine trying to build more workspaces to enabler greater concurrency. I'd be interested to hear if this makes any difference.